1. Field of the Invention
The present invention relates to an image processing device for identifying areas of image data that include character images without relying on a character recognition process.
2. Description of Related Art
The process of separating an area containing only character images (hereinafter called “character image area”) from image data of a document and extract image data consisting only of pixels which constitute character images (hereinafter called “character image data”) has hitherto been carried out by using the character recognition process. However, since the character recognition process depends on the OS (operating system) and has need of a dictionary file, it has not been suitable for being built into a scanner or a dedicated image processing device.
On the other hand, a method of identifying character image areas without relying on the character recognition process is disclosed by JP-A-09-186861. Said method has a problem that, since it makes a judgment whether an extracted area is a character image area based on a premise that the peripheral areas of a character image is monochromatic and most of its frequency components consist of direct current components, characters on colored areas reproduced in screens, for example, such as on magazine pages have peripheral areas that their frequency characteristic lacks said features thus making said method unsuitable for properly extracting character image areas. Moreover, since the method identifies an area by a unit of specified block (8×8 DCT transformation), so that if it makes a misjudgment, it can end up generating a partially defective character, or a strange-looking image when placed in a string of characters.
On the other hand, in an image processing of document image data, it is not necessary to extract a document image area as an area containing only character images but rather it is sufficient to extract it as an area containing mainly character images (hereinafter called “character image containing area”), and an image processing to a document image data can be efficiently and simply carried out, in some cases, by applying an image processing suitable for character images to the character image containing areas thus obtained.